Compute Market Index

A factor model regression on $/GPU-hr across cloud providers. Live prices and a scenario sandbox for negotiating compute.

Index as of Jul 16, 2026

Not an OTC-regulated index.

The compute index

Listings-weighted, chain-linked composite of per-GPU price moves across tracked datacenter GPUs. Base 100 = Mar 12.

115.69 index ▲ 15.7% since Mar 12

100.0 105.0 110.0 115.0 Mar 12 Apr 12 May 14 Jun 14 Jul 16

Model fit R² 0.821 on 52,446 listings.

Per-GPU detail

Daily scarcity, typical hourly price, and how providers price it for each tracked GPU.

GPU

A10 A100 A40 A4000 A5000 A6000 B200 H100 H200 L4 L40 SRTX4090 RTX6000 Ada RTXPRO6000 T4 V100 ▼

Latest $/GPU-hr? $0.61

Providers? 4

Listings? 839

Form factor? PCIE

$0.62 $0.64 $0.66 $0.68 $0.70 Mar 12 Apr 12 May 14 Jun 14 Jul 16

Provider dispersion · Jul 16, 2026

Provider

$0.50 $1.00 $1.50 $2.00 $2.50

Median P25–P75 N

Vast.ai

$0.55 $0.44–$0.60 26

Shadeform

$0.57 $0.57–$0.57 7

Aws

$0.63 $0.43–$0.92 570

Oracle

$0.89 $0.89–$0.89 1

Azure

$1.25 $0.67–$2.72 240

Each bar spans the middle 50% of that provider's listings on Jul 16, 2026. The dot marks the median. All prices are $/GPU-hr for a plain on-demand rental.

Scenario sandbox

Pick a GPU and dial the knobs the model knows about to see what $/GPU-hr that setup implies on our model.

GPU

A10 A100 A40 A4000 A5000 A6000 B200 H100 H200 L4 L40 SRTX4090 RTX6000 Ada RTXPRO6000 T4 V100 ▼

Service level?

Best effort Standard Enterprise

Commitment?

On demand 3 mo 6 mo 1 yr 3 yr 5 yr

Cluster size?

1 2 4 8

Electricity? 16.8¢/kWh

8¢ Cheap hydro 12¢ US industrial 16.8¢ US average 22¢ EU industrial 30¢ EU peak Custom ¢/kWh

Preemptible? Off (stable)

A10 · scenario $/GPU-hr

$0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9/hr

Starting rate? $0.61/hr

Change vs. starting -0.2%

Model estimate, not a live quote. GPU hardware specs stay tied to the GPU you picked; only the scenario knobs above move the price.